Semantic Communication Based on the Coexistence of Knowledge Graph and Large Language Model
Zhiyue Zhang, Wenzhe Lv, Junsheng Mu · 2023
As the wireless communication performance has approached the Shannon limit, semantic communication (SemCom) has attracted much attention as a new communication paradigm to break the channel capacity constraint. To accurately transmit semantics, knowledge graph (KG) has been used for semantic extraction and information recovery at the receiver. Traditional KGs have problems of incomplete structure and high cost, while large language models (LLMs) can help to quickly construct and expand KG. By adopting a symbiotic way of KG and LLM, they can enhance each other and jointly optimize the whole process of SemCom. In this paper, we introduce the basic concepts, basic architecture, and mathematical models of SemCom, KG and LLM, and then discuss three co-optimization strategies for LLMs and KG coexistence to explore the synergistic gains. Finally, we discuss some open issues.